Introduction

Paediatric brain tumour survival rates vary between tumour types and with grade¹. Early identification of those at higher risk may allow better treatment decisions to be made. Conventional descriptors of paediatric brain tumour survival include tumour type and grade, which rely on histopathology being obtained following surgery / biopsy². This is invasive and may not be available due to tumour location or results being delayed. Dynamic susceptibility-contrast (DSC-) MRI involves rapid scanning following injection of a contrast agent and provides non-invasive estimates of perfusion parameters such as relative cerebral blood volume (rCBV). Leakage of contrast agent occurs in low-grade tumours resulting in under- or overestimation of rCBV. We have previously shown that leakage correction is essential in this patient group³. The aim of this work was to assess whether leakage-corrected DSC-MRI parameters could predict survival in paediatric brain tumour patients.

Methods

85 patients underwent pre-treatment DSC-MRI scans at 4 centres on 6 different scanners. Scanning protocols were variable (Table 1). Pixel-by-pixel contrast agent concentration time courses were analysed using the Boxerman model of leakage correction⁴. Estimates of uncorrected and leakage-corrected rCBV (rCBV_uncorr and rCBV_corr, respectively) and the leakage parameter, K₂, were obtained. Patients subsequently underwent surgery / biopsy and tumours were classified and graded. Further treatment was undertaken according to the clinical needs of the patient. Patients were followed up locally. Kaplan Meier survival analysis was performed on patient characteristics including sex, tumour type, grade and DSC-MRI parameters.

Centre	Scanner	B0	Sequence	TR / TE (ms)	TE (ms)	Slice thickness (mm)	No. slices	No. dyn.	Matrix	Pre-bolus?	n
1	Siemens	1.5T	GE-EPI	1490–1643	40	5.0	19–21	60	96 x 96	Y	8
1	Philips	3T	GE-EPI	1830–1865	40	3.5	30	60	96 x 96	Y	3
2	Siemens	3T	GE-EPI	1570	29	3.5	16	60	64 x 64	N	12
3	Philips	3T	GE-EPI	1666–2343	40	4.0	25-35	40	128 x 128	Y	26
4	Philips	1.5T	sPRESTO	16.7-17.2	25	3.5	30-36	60	64-80 x 64-80	Y (n=12) N (n=15) n/k (n=9)	15
4	Philips	3T	sPRESTO	15.5-16.0	24	3.5	30–34	60	128 x 128		16
4	Philips	3T	GE-EPI	582-1866	18.4-40	3.5-7.0	30	40-60	96 x 96		5

Results

Median follow-up time was 88 months (range = 45 – 183 months). At analysis, 20 patients had died, 65 were alive (Table 2). Figure 1 shows example parameter maps from two patients with high-grade tumours. Survival differed between the two patients and was predicted better by median DSC-MRI parameters than by either tumour diagnosis or grade. Across the population, low rCBV_uncorr and rCBV_corr were associated with significantly better survival (p<0.001 and 0.004, respectively) using cut-offs of 2.5 and 2.1, respectively. A positive K₂, associated with leakage correction in low-grade tumours, was predictive of significantly better overall survival (p=0.020). Other prognostic factors were tumour type (p<0.01), tumour grade (p<0.01). Tumour volume, age, sex and centre were not significantly associated with survival.

Parameter	All	Alive	Dead	p-value
n	85	65	20
Median survival (mths)	70.2	88.1	22.2	<0.001**
Tumour vol (cm3)	28.1 ± 26.6	25.7 ± 25.9	36.1 ± 28.0	<0.001**
rCBVuncorr	1.19 ± 4.41	0.65 ± 4.64	2.94 ± 3.05	<0.001**
rCBVcorr	2.13 ± 1.56	1.90 ± 1.31	2.90 ± 2.05	<0.001**
K2	0.009 ± 0.036	0.013 ± 0.039	-0.003 ± 0.022	0.05

Discussion

Patients with well-perfused tumours prior to treatment had significantly worse overall survival than poorly-perfused tumours. rCBV_uncorr was the best parameter at predicting survival. The leakage parameter, K₂, was also significantly predictive of survival with positive K₂ associated with better overall long-term survival. High perfusion is indicative of angiogenesis known to be important for tumour growth and results agree with those presented in other studies^5,6. The significance of DSC-MRI parameters as predictors of survival is in line with conventional survival predictors, such as tumour type and grade. DSC-MRI parameters are available immediately after imaging and are obtained non-invasively. They may be particularly useful for survival prediction in tumours where histological diagnosis may be unavailable or delayed, allowing identification of patients whose prognosis is poor and who may benefit from additional treatments.

Conclusion

In summary, perfusion MRI at diagnosis can aid in predicting which patients are likely to have better survival rates. Low perfusion was associated with better survival. This finding is robust across multiple centres, despite using multiple DSC-MRI protocols.

Acknowledgements

Funded by: CRUK, EPSRC, MRC, NIHR, BCHRF, HHHO

References

¹National Cancer Registration and Analysis Service http://www.ncin.org.uk/cancer_type_and_topic_specific_work/cancer_type_specific_work/cancer_in_children_teenagers_and_young_adults/ Accessed 2021; ²Manias K, et al. Eur J Cancer (2017), 72: 251-65; ³Withey S, et al. Ped Radiol (In press); ⁴Boxerman JL, et al. AJNR (2006), 27(4): 859-67; ⁵Hipp, Neuro Oncol (2011), 13(8): 904-9; ⁶Tensaouti F, et al. Br J Radiol (2016), 89(1066): 20160537.